Your editor took out some time to get back to the lab table and look into that question of ‘thermal memory’ that has been named as possible cause for audible differences between otherwise identically-measuring amplifiers. Specifically, Jan looked at the changes in class A and AB output stage bias current resulting from program dynamics, which are too short to be corrected by the usual thermal compensation loop. Using special-purpose test equipment, it was found that indeed bias current does increase during a high level program burst, and then again decreases during soft passages. So called ‘thermal trak’ transistors, although effective for stabilizing the bias current long-term, were found to be not effective on the time scale investigated. However, changes were found to be limited to a few percent and are unlikely to be audible.
Thermal transient variation of power amp quiescent current - Instrumentation and findings -Jan Didden